Page 100 - Read Online

Page 6 of 26 Li et al. Cancer Drug Resist. 2025;8:31

We conducted gene set enrichment analysis (GSEA) of the differentially expressed genes associated with
these subgroups to explore their biological characteristics. GSEA is frequently used to evaluate changes in
pathways and biological processes within expression datasets . We used the clusterProfiler [27,28] R package to
[38]
conduct and visualize GSEA, using a significance threshold with corrected P-values of <​ 0.05.

Validation of prognostic prediction models at the single-cell level
To validate the prognostic prediction models at a single-cell level, deconvolution was performed on samples
from the TCGA-PRAD to assess the proportion of cancer cell subtypes in each sample. The R package
CIBERSORT , which uses support vector regression, was used to infer the composition of various cell
[15]
subtypes in the tissue samples. A gene expression profile was constructed for CIBERSORT using
differentially expressed genes from each cell subpopulation, which reflected the biological characteristics of
each cell subpopulation, thereby maximizing the reliability and accuracy of the results.

Subsequently, all samples were automatically categorized based on the percentage of high-risk cell
subpopulations within each sample using the Surv cutoff function. This method separates samples into
groups reflecting different levels of high-risk cell subpopulations, with higher levels indicating an increased
risk. This grouping is representative of the definition of risk at the single-cell level. Stratified survival analysis
was conducted on high- and low-risk cell subgroup samples and the KM survival curves were plotted to
validate the model at the single-cell level. Moreover, the ROC curves were generated to assess the predictive
accuracy of high-risk cell subgroup composition for 1- and 5-year survival rates.

Drug resistance analysis
To describe differences in drug resistance, especially that associated with the current first-line treatments,
among various cell subtypes, we used the R package oncoPredict , applied to the Genomics of Drug
[39]
Sensitivity in Cancer (GDSC) database . Significant differences in drug resistance among subgroups were
[40]
detected, and combined box plots and violin plots were used for visualization. Statistical significance was
determined using the Wilcoxon rank-sum test. The results were presented as a bubble plot.

Cell culture
Human prostate hyperplasia cell line BPH-1, PCa cell lines PC-3 and 22Rv1 were purchased from American
Type Culture Collection (ATCC). All cells were cultured in RPMI-1640 medium supplemented with 10%
fetal bovine serum. The cell lines were maintained at 37 °C in a 5% CO environment. Furthermore, these cell
2
lines were cultured for no more than 20 generations and subjected to routine testing to confirm their absence
of mycoplasma contamination.

Real-time quantitative polymerase chain reaction
Total RNA extraction was extracted using the RNAsimple Total RNA Kit (TIANGEN). The extracted total
RNA was dissolved in RNase-free water. cDNA synthesis was performed using the RevertAid First Strand
cDNA Synthesis Kit (ThermoFisher) and stored at -20 °C. For quantitative PCR, cDNA was used with
SuperReal PreMix Plus SYBR Green Supermix (TIANGEN) in the LightCycler 480 Real-Time PCR System
(Roche) following the manufacturer’s instructions. Fluorescence signals were recorded, and β-actin was used
as the reference gene. Relative expression levels were analyzed using the 2 -ΔΔCt method. The primer
sequences are provided in Supplementary Table 1.

Western blot
Cells were collected, lysed using a lysis buffer, and then centrifuged to collect the supernatant. The
supernatant was heated to 95 °C for 5 min. Protein samples were separated by SDS-PAGE electrophoresis
and semi-dry transferred onto an NC membrane (Millipore). The membrane was blocked in Tris-buffered

93